add test demonstrating problems with json sampler:

- the metric set content gets mixed
- the metric set has only the schema of the first message seen
- messages get lost

ldms/scripts/examples/json_stream_sampler

@@ -0,0 +1,96 @@
+export plugname=json_stream_sampler
+export dsname=$(ldms_dstat_schema_name mmalloc=1 io=1 fd=1 stat=1 auto-schema=1)
+export dstat_schema=$dsname
+export LDMSD_LOG_LEVEL=ERROR
+export LDMSD_LOG_TIME_SEC=1
+export LDMSD_EXTRA="-m 128m"
+
+# Defining and using fakepid to allow testing without root & ldms-notify.
+# good enough for sorting storage pipeline issues.
+# each gets a new msgno so we can check for missing data
+# fake_pid  <daemon number to post to> <pid to fake> <schema name> <event name> <stream name>
+# uses specific formats for events task_init_priv task_exit
+fmsgno=0
+function fake_pid {
+	((fmsgno++))
+	fdaemon=$1
+	fpid=$2
+	fschema=$3
+	fevent=$4
+	fstream=$5
+	case $fevent in
+	task_init_priv)
+		fmsg="{\"msgno\":$fmsgno,\"schema\":\"$fschema\",\"event\":\"task_init_priv\",\"timestamp\":1731000790,\"context\":\"*\",\"data\":{\"start\":\"1729013913.013913\",\"job_id\":\"0\",\"serial\":1,\"os_pid\":$fpid,\"uid\":4294967295,\"gid\":4294967295,\"is_thread\":0,\"exe\":\"/usr/lib/systemd/systemd\",\"start_tick\":\"8\",\"task_pid\":1,\"task_global_id\":-1}}"
+		;;
+	task_exit)
+		fmsg="{\"msgno\":$fmsgno,\"schema\":\"$fschema\",\"event\":\"task_exit\",\"timestamp\":1731000792,\"context\":\"*\",\"data\":{\"start\":\"1731000777.000777\",\"job_id\":\"0\",\"serial\":1131,\"os_pid\":$fpid,\"uid\":4294967295,\"gid\":4294967295,\"is_thread\":0,\"exe\":\"/usr/lpp/mmfs/bin/mmksh\",\"duration\":15.161968946456909,\"start_tick\":\"198686390\",\"task_pid\":140962,\"task_global_id\":-1}}"
+		;;
+	*)
+		fmsg="{\"msgno\":$fmsgno,\"schema\":\"$fschema\",\"event\":\"$fevent\",\"timestamp\":1731000792,\"data\":{\"start\":\"1731000777.000777\",\"job_id\":\"0\",\"os_pid\":$fpid,\"uid\":4294967295,\"gid\":4294967295,\"exe\":\"/mypath/myprog\",\"start_tick\":\"198686390\"}}"
+		;;
+	esac
+	if test -n "$PORT"; then
+		iport=$PORT
+	else
+		iport=${ports[$fdaemon]}
+	fi
+	echo sending "<$fmsg>" to $iport
+	echo "$fmsg" | ldmsd_stream_publish -p $iport -a none -s $fstream -t json -x sock -h localhost
+}
+
+portbase=61070
+rm -f $LOGDIR/json*.log
+rm -rf $STOREDIR
+mkdir -p $STOREDIR
+VGARGS="--tool=drd --suppressions=/scratch1/baallan/ovis/ldms/scripts/examples/linux_proc_sampler.drd.supp"
+VGARGS="--leak-check=full --track-origins=yes --trace-children=yes --show-leak-kinds=definite --time-stamp=yes --keep-debuginfo=yes"
+DAEMONS $(seq 3)
+#vgon
+LDMSD 1
+vgoff
+#vgon
+LDMSD 2
+vgoff
+LDMSD 3
+vgoff
+SLEEP 1
+
+fake_pid  1 2000 linux_task_data task_init_priv ldms-notify
+fake_pid  1 2002 slurm_task_data task_init_priv ldms-notify
+fake_pid  1 1999 lsf_task_data   task_init_priv ldms-notify
+SLEEP 1
+fake_pid  1 1999 lsf_task_data   task_exit      ldms-notify
+fake_pid  1 2002 slurm_task_data task_exit      ldms-notify
+fake_pid  1 2000 linux_task_data task_exit      ldms-notify
+SLEEP 1
+fake_pid  1 2001 myschema        myevent        ldms-notify
+fake_pid  1 2002 myschema        myevent        ldms-notify
+
+SLEEP 1
+for lc in $(seq 3); do
+	echo "CHECKING STATUS ON $lc:"
+	ldmsd_controller --auth none --port ${ports[$lc]} --cmd stream_status
+done
+SLEEP 2
+MESSAGE ldms_ls on host 1:
+LDMS_LS 1 -v
+MESSAGE ldms_ls on host 2:
+SLEEP 1
+LDMS_LS 2 -v
+SLEEP 5
+MESSAGE stream_client_dump on sampler daemon 1
+for lc in $(seq 2); do
+	echo "CHECKING CLIENT DUMP ON $lc:"
+	ldmsd_controller --auth none --port ${ports[$lc]} --cmd stream_client_dump
+	SLEEP 1
+done
+SLEEP 5
+for lc in $(seq 3); do
+	LDMS_LS $lc -v
+	SLEEP 2
+done
+SLEEP 2
+KILL_LDMSD 3 2 1
+file_created $STOREDIR/node/$dsname
+rollover_created $STOREDIR/blobs/ldms-notify.DAT
+# check here for csv and line count of csv for stream messages expected

ldms/scripts/examples/json_stream_sampler.1

@@ -0,0 +1,17 @@
+load   name=json_stream_sampler
+config name=json_stream_sampler producer=localhost${i} instance=localhost${i}/jss component_id=${i} stream=ldms-notify perm=0644 heap_sz=64m
+start  name=json_stream_sampler interval=1000000 offset=0
+#log sampler.json_stream DEBUG
+
+load name=dstat
+config name=dstat producer=localhost${i} instance=localhost${i}/${dstat_schema} component_id=${i} mmalloc=1 io=1 fd=1 auto-schema=1 stat=1 perm=777
+start name=dstat interval=1000000 offset=0
+
+# define updater to push sets to L1 as they appear
+##updtr_add name=push_pidset push=onchange interval=1000000
+# catch self in .*
+##updtr_prdcr_add name=push_pidset regex=.*
+# match sets by schema name
+##updtr_match_add name=push_pidset regex=linux_task_data match=schema
+# begin pushing
+##updtr_start name=push_pidset

ldms/scripts/examples/json_stream_sampler.2

@@ -0,0 +1,44 @@
+load name=dstat
+config name=dstat producer=localhost${i} instance=localhost${i}/${dstat_schema} component_id=${i} mmalloc=1 io=1 fd=1 auto-schema=1 stat=1 perm=777
+start name=dstat interval=1000000 offset=0
+
+# blobs must be allowed by writer plugin and prdcr_subscribe by daemon
+load name=blob_stream_writer plugin=blob_stream_writer
+config name=blob_stream_writer path=${STOREDIR} container=blobs stream=ldms-notify types=1
+
+prdcr_add name=localhost1 host=${HOST} type=active xprt=${XPRT} port=${port1} reconnect=2000000
+prdcr_subscribe regex=.* stream=ldms-notify
+prdcr_start name=localhost1
+
+updtr_add name=allhosts interval=1000000 offset=100000
+updtr_prdcr_add name=allhosts regex=.*
+updtr_start name=allhosts
+
+load name=store_csv
+config name=store_csv path=${STOREDIR} altheader=0
+
+strgp_add name=json_stream_store_linux plugin=store_csv schema=linux_task_data container=pnode
+strgp_prdcr_add name=json_stream_store_linux regex=.*
+strgp_start name=json_stream_store_linux
+
+strgp_add name=json_stream_store_lsf plugin=store_csv schema=lsf_task_data container=pnode
+strgp_prdcr_add name=json_stream_store_lsf regex=.*
+strgp_start name=json_stream_store_lsf
+
+strgp_add name=json_stream_store_slurm plugin=store_csv schema=slurm_task_data container=pnode
+strgp_prdcr_add name=json_stream_store_slurm regex=.*
+strgp_start name=json_stream_store_slurm
+
+strgp_add name=json_stream_store_my plugin=store_csv schema=myschema container=pnode
+strgp_prdcr_add name=json_stream_store_my regex=.*
+strgp_start name=json_stream_store_my
+
+
+# define updater to push sets to L1 as they appear
+#updtr_add name=push_pidset push=onchange interval=1000000
+# catch self in .*
+#updtr_prdcr_add name=push_pidset regex=.*
+# match sets by schema name
+#updtr_match_add name=push_pidset regex=linux_task_data match=schema
+# begin pushing
+#updtr_start name=push_pidset

ldms/scripts/examples/json_stream_sampler.3

@@ -0,0 +1,17 @@
+load name=blob_stream_writer plugin=blob_stream_writer
+config name=blob_stream_writer path=${STOREDIR} container=blobs_L2 stream=ldms-notify types=1
+
+prdcr_add name=localhost2 host=${HOST} type=active xprt=${XPRT} port=${port2} interval=2000000
+prdcr_subscribe regex=.* stream=ldms-notify
+prdcr_start name=localhost2
+
+updtr_add name=allhosts interval=1000000 offset=200000
+updtr_prdcr_add name=allhosts regex=.*
+updtr_start name=allhosts
+
+load name=store_csv
+config name=store_csv path=${STOREDIR} altheader=0
+
+strgp_add name=store_dstat plugin=store_csv schema=${dstat_schema} container=node
+strgp_prdcr_add name=store_dstat regex=.*
+strgp_start name=store_dstat

ldms/src/sampler/blob_stream/Plugin_blob_stream_writer.man

@@ -48,6 +48,10 @@ timing=1
 .br
 Enable writing timestamps to a separate file.
 .TP
+types=1
+.br
+Enable writing message types to a separate file.
+.TP
 spool=1
 .br
 Move closed files to the directory <path>/<container>/spool/.
@@ -114,6 +118,13 @@ Each timestamp is written to the .TIMING file as a binary pair (tv_sec, tv_usec)
 with each value stored as a little-endian 64 bit value which should be
 read and then converted with le64toh.
 
+Optionally (if types=1 given) the additional file
+$path/$container/$stream.TYPE.$create_time
+is created containing binary type characters corresponding to the messages.
+The TYPE file begins with an 8 byte magic number: blobtyp\\0.
+Each type is a character (j: json, s:string(fixed length as determined by corresponding offset).
+There is no separator between message type indicators.
+
 .SH NOTES
 .PP
 This writer is in development and may be changed at any time.

ldms/src/sampler/json/Plugin_json_stream_sampler.man

@@ -18,7 +18,7 @@ Within ldmsd_controller or a configuration file:
 .OP uid=\fIUID\fR
 .OP gid=\fIGID\fR
 .OP perm=\fIPERM\fR
-.OP heap_szperm=\fIBYTES\fR
+.OP heap_sz=\fIBYTES\fR
 .YS
 
 .SH DESCRIPTION